Skip to content

Wafer-scale epitaxial growth of AlN (0001) on heteroepitaxial diamond (111)

FieldValue
JournalApplied Physics Letters
Published2026-02-16
AuthorsJürgen Weippert, Balasubramanian Sundarapandian, Mohit Raghuwanshi, Mario Prescher, Lutz Kirste
DOI10.1063/5.0302996

Wafer-scale wurtzite AlN (0001) grew epitaxially on heteroepitaxial diamond (111). The best AlN 0002 rocking curve FWHM was 1.09°, with a surface roughness Rq of 1.5 nm on as-grown diamond.

The process used a 50.8 mm Si (111) substrate and produced a 100 nm AlN layer. XRD and AFM showed a closed layer without cracks or unwanted grain orientations.

The complete stack ran from AlN through heteroepitaxial diamond, Ir, YSZ, and Si (111). Both seed layers were deposited by magnetron sputtering epitaxy. YSZ was 45 nm thick, and Ir was 140 nm thick. The Si wafers carried 4° or 6° off-cuts toward (011). The authors reused deposition parameters reported in their earlier work.

XRD phase analysis of Ir/YSZ/Si showed only the desired material orientation. The Ir 111 rocking curve was narrower than 0.6° FWHM. An Ir 200 pole figure showed three main epitaxial spots. Twinned material contributed about 2% of the integrated XRD intensity. The YSZ roughness Rq was 0.5 nm on both samples. Ir roughness was 0.7 nm for the 4° off-cut and 1.1 nm for the 6° off-cut.

AFM showed sharp YSZ steps, occasional rotated grains, and step-bunched Ir. YSZ measurements used tapping mode, a 42 N/m NCHR tip, and cleanroom ambient conditions. Earlier diamond wafers had been reported up to 4 in. diameter in the (001) orientation. The required (111) orientation had not been available as a wafer for AlN (0001) growth. This work produces that orientation on 50.8 mm Ir/YSZ/Si (111) substrates.

Diamond nucleated by bias-enhanced nucleation in a 2.45 GHz ellipsoid PECVD reactor. The two samples differed in off-cut, thickness, reactor cooling, oxygen exposure, and final surface treatment.

SampleDiamond surfaceGrowth time and rateThicknessDiamond 111 FWHM
A4° off-cut, polished690 h at 0.3 μm/h230 μm1.24°
B6° off-cut, as grown135 h at 70 nm/h10 μm0.26°

Sample A used 3.8 kW at 970 K and 6.5 kPa. Sample B used 4.2 kW at 960 K and 8.0 kPa. CH4/H2 was 1.6% for A, with O2/H2 at 1.6%. For B, CH4/H2 was 1.3%, with no added oxygen. Continuous oxygen addition etched sample A during growth. Sample A’s reactor plate was largely decoupled from the environment. Sample B’s reactor had active plate cooling.

SEM at 5 kV showed mostly triangular surface features on sample B. Sample A contained small {100} twins, visible as rectangles or pyramids. The diamond 004 pole figure had three expected spots and few side signals. Its 119.5° detection angle corresponded to diamond 004. AFM and XRD identified sample B as the higher-quality diamond condition.

Before AlN deposition, sample A was polished to Rq of 4 nm and cleaned with sulphonitric acid. Sample B remained as grown. Both samples used identical process flows after their respective surface preparations. AlN was deposited in an Evatec Clusterline 200 system using magnetron sputtering epitaxy.

ICP etching removed surface terminations at 500 W ICP power and 50 W RF power. The etch used 0.15 Pa Ar for 60 s. AlN growth used 3 kW pulsed DC from a 300 mm Al target. The substrate temperature was 1050 K, pressure was 0.2 Pa, and growth lasted 120 s. The target-substrate distance was 60 mm. The deposition atmosphere was pure nitrogen. The resulting AlN layer was 100 nm thick.

XRD phase analysis, rocking curves, and pole figures tested phase and orientation. SEM and AFM tracked surface morphology and roughness.

SEM showed that the AlN morphology followed mesoscopic features in the diamond. This effect was strongest on as-grown sample B. Surface SEM used a 20° viewing angle. The cleaved edge, viewed at 45°, showed columnar AlN growth. Surface grains measured about 50 nm in diameter.

The off-cut direction was not resolved on sample B because its diamond terraces appeared planar and (111)-oriented. AFM gave an AlN Rq of 0.8 nm on sample A. Sample B reached Rq of 1.5 nm after mesoscopic features were excluded. Ribbons on sample A followed the diamond off-cut direction.

The authors concluded that both samples formed a uniform, closed AlN layer without abnormal island or crack formation. XRD found no undesired AlN misorientation on either sample. Both pole figures contained six spots, as expected for epitaxial growth.

On sample B, phase analysis at the wafer center and 5 mm from the edge showed the desired orientations throughout the stack. AlN 0002 rocking curves were 1.09° and 1.15° at those two positions. Sample A produced a broader 1.9° to 4.0° range across its measurement points. The reported FWHM values use the 0002 reflection, although the epitaxial plane in the title is AlN (0001).

The authors judged AlN on sample B comparable in quality to AlN grown on Si (111) by a similar process. Sample A had lower AlN crystal quality.

Off-cut, diamond thickness, crystal quality, reactor conditions, and surface treatment differed between samples. The present data cannot assign the AlN difference to a single cause. Further preparation is needed to separate diamond quality from polishing effects.

ICP etching before AlN deposition remains an open variable. Prior work suggested that uneven etching may affect the interlayer. Earlier monolithic epitaxy used small HPHT crystals rather than wafer-scale layers.

Earlier bonding approaches were limited to millimeter-sized substrates and sensitive to sample warping. The nanoseeded diamond concept produced polycrystalline diamond, which the authors say does not reach full thermal potential. The heteroepitaxial route therefore tests a different integration strategy. The paper reports no electrical, thermal, or device-level measurements for the combined stack.

The measured result is a 100 nm AlN seed layer for further nitride overgrowth. The authors propose MOCVD growth of other nitride materials to complete the HEMT stack. Device fabrication and reliability tests under high-power operation would then follow.

The authors state that inefficient cooling raises device temperature, which can increase resistivity and shorten lifetime. Diamond’s cited thermal conductivity reaches 2300 W/(m K). The paper compares this with 350 W/(m K) for V-doped 4H-SiC and 150 W/(m K) for silicon. These reference values motivate diamond as a heat spreader for power electronics and radio-frequency technology.

The paper did not measure the thermal conductivity of the grown diamond layer. It also did not quantify heat removal, electrical performance, or efficiency. Uniform AlN coverage and low roughness were identified as essential requirements for HEMT fabrication. For material selection, the paper requires a (111) diamond surface and treats surface condition as an unresolved process variable. The result supports further HEMT development, but device performance remains unproven in this work.

The paper needs a wafer-scale single-crystal diamond layer with a confirmed (111) surface for epitaxial AlN (0001) growth. 6CCVD’s diamond materials catalog includes single-crystal MPCVD diamond on plates and wafers up to 125 mm, although the paper used PECVD and does not confirm a catalog-compatible (111) orientation. The engineering support page can be used to check orientation, surface preparation, and custom integration requirements.

View Original Abstract

Wurtzite-type aluminum nitride (AlN) films were deposited by magnetron sputtering epitaxy (MSE) on wafer-scale diamond layers, which in turn were grown heteroepitaxially by plasma-enhanced chemical vapor deposition applying bias-enhanced nucleation on an Ir seed layer. The seed was fabricated on Si (111) wafers with an yttria-stabilized zirconia buffer layer by MSE. Analysis by x-ray diffraction has confirmed fully epitaxial growth throughout the entire stack. Due to diamond’s high thermal conductivity, which makes it an effective heat sink, this stack constitutes an important step toward monolithically integrated high-efficiency III-nitride-diamond power electronics.

  1. 2007 - CRC Handbook of Chemistry and Physics